Crystal D. Linkletter

Variable selection for Gaussian process models in computer experiments

In many situations, simulation of complex phenomena requires a large number of inputs and is computationally expensive. Identifying the inputs that most impact the system so that these factors can be further investigated can be a critical step in the scientific endeavor. In computer experiments, it is common to use a Gaussian spatial process to model the output of the simulator. In this article we introduce a new, simple method for identifying active factors in computer screening experiments. The approach is Bayesian and only requires the generation of a new inert variable in the analysis; however, in the spirit of frequentist hypothesis testing, the posterior distribution of the inert factor is used as a reference distribution against which the importance of the experimental factors can be assessed. The methodology is demonstrated on an application in material science, a computer experiment from the literature, and simulated examples.

Adult Health Behaviors Over the Life Course by Sexual Orientation

OBJECTIVES We estimated differences in health behaviors among adults by sexual orientation. METHODS We pooled 4 years of data (2001, 2003, 2005, and 2007) from the California Health Interview Survey. We estimated the frequency of smoking, alcohol use, healthy dietary behaviors, physical activity, and health care utilization, and we used logistic regression modeling to determine the odds of each behavior with increasing age and for 2 age groups: younger than 50 years and 50 years old or older. RESULTS At any adult age, lesbians had greater odds of smoking and binge drinking than did heterosexual women, and gay and bisexual men had greater health care utilization than did heterosexual men. Other risk behaviors differed with age. CONCLUSIONS Some behavioral change interventions should target lesbians, gays, and bisexuals at all ages, whereas other interventions should specifically target individuals at younger ages.

Short Sleep Duration Is Associated With Carotid Intima-Media Thickness Among Men in the Coronary Artery Risk Development in Young Adults (CARDIA) Study

Background and Purpose— Carotid intima-media thickness (CIMT) is a subclinical marker of cardiovascular disease. Recent studies suggest that shorter sleep duration is a risk factor for cardiovascular disease, but there is limited evidence regarding this association using high-quality, objective assessments of sleep. The aim of this study is to determine whether sleep duration is associated with CIMT. Methods— The study used an observational cohort consisting of 617 black and white middle-aged healthy participants (37–52 years; 58% female) in the Coronary Artery Risk Development in Young Adults (CARDIA) Study. Multivariable-adjusted linear regression analyses were performed. Sleep duration was measured using wrist actigraphy monitors. CIMT was calculated using the average of 20 measurements of the mean common carotid, bulb, and internal CIMT, which was assessed using ultrasound images. Results— After adjusting for covariates, 1 hour of longer sleep duration was associated with 0.026 mm less CIMT among men (P=0.02; 95% CI, −0.047 to −0.005) and 0.001 mm less CIMT among women (P=0.91; 95% CI, −0.020 to 0.022). Segment-specific analyses indicated that the carotid bulb was a key driver of the observed association. Conclusions— Shorter objectively assessed sleep duration was associated with greater CIMT among men but not women.

Spatio-Spectral Mixed-Effects Model for Functional Magnetic Resonance Imaging Data

The goal of this article is to model cognitive control related activation among predefined regions of interest (ROIs) of the human brain while properly adjusting for the underlying spatio-temporal correlations. Standard approaches to fMRI analysis do not simultaneously take into account both the spatial and temporal correlations that are prevalent in fMRI data. This is primarily due to the computational complexity of estimating the spatio-temporal covariance matrix. More specifically, they do not take into account multiscale spatial correlation (between-ROIs and within-ROI). To address these limitations, we propose a spatio-spectral mixed-effects model. Working in the spectral domain simplifies the temporal covariance structure because the Fourier coefficients are approximately uncorrelated across frequencies. Additionally, by incorporating voxel-specific and ROI-specific random effects, the model is able to capture the multiscale spatial covariance structure: distance-dependent local correlation (within an ROI), and distance-independent global correlation (between-ROIs). Building on existing theory on linear mixed-effects models to conduct estimation and inference, we applied our model to fMRI data to study activation in prespecified ROIs in the prefontal cortex and estimate the correlation structure in the network. Simulation studies demonstrate that ignoring the multiscale correlation leads to higher false positive error rates.

Functional connectivity: shrinkage estimation and randomization test.

We develop new statistical methods for estimating functional connectivity between components of a multivariate time series and for testing differences in functional connectivity across experimental conditions. Here, we characterize functional connectivity by partial coherence, which identifies the frequency band (or bands) that drives the direct linear association between any pair of components of a multivariate time series after removing the linear effects of the other components. Partial coherence can be efficiently estimated using the inverse of the spectral density matrix. However, when the number of components is large and the components of the multivariate time series are highly correlated, the spectral density matrix estimate may be numerically unstable and consequently gives partial coherence estimates that are highly variable. To address the problem of numerical instability, we propose a shrinkage-based estimator which is a weighted average of a smoothed periodogram estimator and a scaled identity matrix with frequency-specific weight computed objectively so that the resulting shrinkage estimator minimizes the mean-squared error criterion. Compared to typical smoothing-based estimators, the shrinkage estimator is more computationally stable and gives a lower mean squared error. In addition, we develop a randomization method for testing differences in functional connectivity networks between experimental conditions. Finally, we report results from numerical experiments and analyze an EEG data set recorded during a visually-guided hand movement task.

Fatty Acid desaturase gene variants, cardiovascular risk factors, and myocardial infarction in the costa rica study.

Genetic variation in fatty acid desaturases (FADS) has previously been linked to long-chain polyunsaturated fatty acids (PUFAs) in adipose tissue and cardiovascular risk. The goal of our study was to test associations between six common FADS polymorphisms (rs174556, rs3834458, rs174570, rs2524299, rs174589, rs174627), intermediate cardiovascular risk factors, and non-fatal myocardial infarction (MI) in a matched population based case–control study of Costa Rican adults (n = 1756). Generalized linear models and multiple conditional logistic regression models were used to assess the associations of interest. Analyses involving intermediate cardiovascular risk factors and MI were also conducted in two replication cohorts, The Nurses’ Health Study (n = 1200) and The Health Professionals Follow-Up Study (n = 1295). In the Costa Rica Study, genetic variation in the FADS cluster was associated with a robust linear decrease in adipose gamma-linolenic, arachidonic, and eicosapentaenoic fatty acids, and significant or borderline significant increases in the eicosadienoic, eicosatrienoic, and dihomo-gamma-linolenic fatty acids. However, the associations with adipose tissue fatty acids did not translate into changes in inflammatory biomarkers, blood lipids, or the risk of MI in the discovery or the replication cohorts. In conclusion, fatty acid desaturase polymorphisms impact long-chain PUFA biosynthesis, but their overall effect on cardiovascular health likely involves multiple pathways and merits further investigation.

Peer associations for substance use and exercise in a college student social network.

UNLABELLED Substance use and exercise have opposite trajectories in young adulthood, and research indicates that peers are influential for both of these health behaviors, but simultaneous investigations of peer associations with substance use and exercise have not been conducted. OBJECTIVE Use a college residence hall peer network to examine associations between peer behaviors and alcohol use, marijuana use, and exercise behavior. METHOD 129 undergraduates (51.9% female, 48.1% non-Hispanic White; 84.5% first-year students) in one residence hall completed a Web-based survey of substance use and exercise and identified up to 10 students in the residence hall who were important to them. Two social network analytic methods, community detection cluster analysis and network autocorrelation modeling, were used to identify peer groupings and to examine the associations between peer and participant behaviors, respectively. RESULTS Participants nominated an average of 4.1 residence hall members, and 53.9% of the ties were reciprocal. 6 clusters were identified that differed significantly on demographics, college activities, substance use, and exercise. Weekly volume of alcohol consumed among nominated peers was significantly associated with that of participants, and all other covariates, including gender and athlete status, were not significant. Peer marijuana use also was associated with participant use after controlling for covariates. Exercise levels of nominated peers were not associated with exercise levels of participants. CONCLUSIONS College student networks may be good targets for health-related prevention programs. Programs that use close-proximity peers to influence the behavior of others might be more effective with substance use as the target behavior than exercise.

Age, Antipsychotics, and the Risk of Ischemic Stroke in the Veterans Health Administration

Background and Purpose— Time-dependent effects of antipsychotics on risk of stroke and potential effect modification by age have not been fully investigated. A case–case–time–control design uses within- and between-case comparisons to evaluate short-term effects at the same time as adjusting for unmeasured time-invariant confounders and exposure-time trends. Methods— We conducted a case–case–time–control design study using data from the Veterans Health Administration. Veterans with inpatient hospitalizations for ischemic stroke between 2002 and 2007 were included. For every stroke case, the “current” exposure period was defined as 1 to 30 days before hospitalization and the “reference” period as 91 to 120 days before hospitalization. Exposure during the current period was compared with exposure during the reference period within cases. Exposure-time trend-adjusted estimates of the effect of antipsychotic exposure on risk of stroke were obtained by dividing exposure odds for antipsychotic exposure by average exposure odds for other medications over the same time period among the same cases. Results— After adjusting for exposure-time trends, odds of stroke were 1.8 (95% CI, (1.7–1.9) times higher when exposed to antipsychotics than when unexposed. Age-stratified estimates suggest a greater triggering effect of antipsychotics among older patients. Conclusions— Exposure to antipsychotics may be a proximal trigger for stroke. Elevation in risk is apparent after brief exposure to antipsychotics.

Decreased Births Among Black Female Adolescents Following School Desegregation

Although the socioeconomic impact of school desegregation in the U.S. has been well documented, little is known about the health consequences of this policy. The purpose of this study was to quantify the associations between school desegregation and adolescent births among black and white females. We compared the change in prevalence of adolescent births in areas that implemented school desegregation plans in the 1970s with areas that implemented school desegregation plans in other decades, using difference-in-difference methods with 1970 and 1980 Census microdata. School desegregation policy in the U.S. in the 1970s was associated with a significant reduction of 3.2 percentage points in the prevalence of births among black female adolescents between 1970 and 1980. This association was specific to black female adolescents and was not observed among white adolescents.

Genetic variation in fatty acid elongases is not associated with intermediate cardiovascular phenotypes or myocardial infarction

Background/objectives:Elongases 2, 4 and 5, encoded by genes ELOVL2, ELOVL4 and ELOVL5, have a key role in the biosynthesis of very long chain polyunsaturated fatty acids (PUFAs). To date, few studies have investigated the associations between elongase polymorphisms and cardiovascular health. We investigated whether ELOVL polymorphisms are associated with adipose tissue fatty acids, serum lipids, inflammation and ultimately with nonfatal myocardial infarction (MI) in a Costa Rican population.Subjects/methods:MI cases (n=1650) were matched to population-based controls (n=1650) on age, sex and area of residence. Generalized linear and multiple conditional logistic regression models were used to assess the associations between seven common ELOVL polymorphisms and cardiometabolic outcomes. Analyses were replicated in The Nurses’ Health Study (n=1200) and The Health Professionals Follow-Up Study (n=1295).Results:Variation in ELOVL2, ELOVL4 and ELOVL5 was not associated with adipose tissue fatty acids, intermediate cardiovascular risk factors or MI. In the Costa Rica study, the number of the minor allele copies at rs2294867, located in the ELOVL5 gene, was associated with an increase in total and LDL cholesterol (adjusted P-values=0.001 and <0.0001 respectively). Additionally, the number of the minor allele copies at rs761179, also located in the ELOVL5 gene, was significantly associated with an increase in total cholesterol (adjusted P-value=0.04). However, the observed associations were not replicated in independent populations.Conclusion:Common genetic variants in elongases are not associated with adipose tissue fatty acids, serum lipids, biomarkers of systemic inflammation, or the risk of MI.